CN116208970B - Air-ground collaboration unloading and content acquisition method based on knowledge-graph perception - Google Patents
Air-ground collaboration unloading and content acquisition method based on knowledge-graph perception Download PDFInfo
- Publication number
- CN116208970B CN116208970B CN202310408910.2A CN202310408910A CN116208970B CN 116208970 B CN116208970 B CN 116208970B CN 202310408910 A CN202310408910 A CN 202310408910A CN 116208970 B CN116208970 B CN 116208970B
- Authority
- CN
- China
- Prior art keywords
- representing
- node
- user
- edge
- content
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D30/00—Reducing energy consumption in communication networks
- Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks
Abstract
The invention discloses a space-ground cooperation unloading and content acquisition method based on knowledge-graph perception, which belongs to the technical field of mobile communication and comprises the following steps: in the urban environment with the lower edge covered by the 5G macro base station, constructing a hollow cooperative unloading and content acquisition model of an on-board self-organizing network environment; dynamically arranging and constructing an edge service knowledge graph based on edge node attributes and a plurality of time-varying parameter knowledge relations among nodes; establishing a system utility function weighted by multiple performance indexes; and constructing space-ground collaborative unloading and content acquisition optimal strategies based on edge service knowledge graph perception for different types of user demands by adopting a theoretical optimization and clipping near-end strategy optimization algorithm. The invention solves the problems of space-ground collaboration unloading and content acquisition of users with different service demands in the vehicle-mounted self-organizing network in the edge urban environment, reduces the edge service delay and lease utility expenditure while maximizing arrangement and utilization of the edge idle resources, and improves the service quality demands of the edge users.
Description
Technical Field
The invention belongs to the technical field of mobile communication, and particularly relates to a space-ground collaboration unloading and content acquisition method based on knowledge-graph sensing.
Background
With the development and popularity of intelligent terminals, a number of delay-sensitive and computation-intensive applications, such as face recognition, natural language processing, real-time streaming media, and virtual/augmented reality, are emerging, which present new challenges for edge computing offloading. In order to fully explore the influence of time-varying parameters among edge devices on the service quality of users in edge service and maximize arrangement and utilization of edge idle resources, task unloading and content acquisition strategies are made for different types of users with different requirements in time, and the method has important significance for research on space-ground collaborative unloading and content acquisition based on knowledge-graph perception in a vehicle-mounted self-organizing network.
In recent years, knowledge-centric edge computing architectures have become a research hotspot due to the ability to perceive user-dependent tasks, network structures, or edge-limited resources. However, the high-speed mobility of the vehicles and the increase of the registration number lead to the occurrence of fluctuation and idling of edge resources, the influence of complex logic parameters among heterogeneous edge devices on the edge unloading of the dependent tasks cannot be reflected at the same time, and a large amount of idle resources among the vehicles can not be fully arranged and utilized when a single edge service node is selected for unloading along with the exponential increase of the number of the vehicles on the road. Therefore, how to arrange and construct the dynamic knowledge relationship of the edge service knowledge graph is a key problem of the research of the invention, and meanwhile, the discovery and sharing virtualization of the edge adjacent resources of the Internet of vehicles are of great research value.
The research significance and research value of dynamic arrangement of multi-time-varying parameter logic knowledge relation between edge environments and discovery sharing of edge adjacent resources are comprehensively improved, and a space-to-ground collaborative unloading and content acquisition method based on knowledge graph perception is needed to better optimize task unloading and content acquisition of edge users and improve service quality experience of the edge users.
Disclosure of Invention
In order to solve the problems, the invention provides an air-ground collaboration unloading and content acquisition method based on knowledge graph perception, which aims at the dependence task unloading requirements and the different content acquisition requirements of different types of users, and solves the problems of air-ground collaboration unloading and content acquisition of different types of users in the vehicle-mounted self-organizing network edge city environment by taking into consideration the dynamic arrangement of a plurality of time-varying parameter knowledge between vehicle edge environment equipment, the perception deployment of air edge nodes and the establishment of a vehicle virtual shared resource pool in the edge nodes and the users.
The technical scheme of the invention is as follows:
a space-ground collaboration unloading and content acquisition method based on knowledge-graph perception comprises the following steps: step 1, constructing an edge vehicle-mounted self-organizing network environment hollow cooperative unloading and content acquisition model consisting of a macro base station, remote cloud, non-preset track aerial edge nodes, road mobile vehicles and two types of mobile users with different types and different requirements in a 5G macro base station coverage lower edge urban environment; step 2, dynamically arranging and constructing an edge service knowledge graph based on edge node attributes and a plurality of time-varying parameter knowledge relations among nodes; step 3, analyzing user service quality influence factors under different service strategies and establishing a multi-performance index weighted system utility function according to node perception deployment and a vehicle-mounted self-organizing network shared resource pooling knowledge model; and 4, constructing an air-ground collaborative unloading and content acquisition optimal strategy based on edge service knowledge graph perception for different types of user demands by adopting a theoretical optimization and clipping near-end strategy optimization algorithm.
Further, in the step 1, an edge city air-ground cooperation scene of a 5G macro base station covering remote cloud, an air edge node, road moving vehicles and two types of mobile users with different types and different requirements is considered; the macro base station is provided with an edge knowledge server for knowledge spectrum sensing, senses physical entity information of users, vehicles and aerial edge nodes in a coverage area, is connected with a remote cloud in an optical fiber communication mode, and contains a large amount of computing resources and cache content fragment sets; meanwhile, quantitatively labeling the positions of different physical entities in the scene by establishing a three-dimensional coordinate system; the unmanned aerial vehicle with fixed height flies at a non-preset angle to be used as an air edge node for flexible perception decision deployment, a minimum safety distance is set between adjacent air edge nodes, and the air edge node contains computing resources and part of the existing cache content set updated periodically so as to assist a user to carry out task unloading, relay forwarding and cache content providing, wherein the air edge node and a macro base station communicate through an air link; two kinds of task users with different moving speeds and tolerant time delays exist in the scene, namely a common user and a vehicle-mounted user; a plurality of moving vehicle nodes exist on the surrounding roads of the cell, each vehicle node carries a virtual machine capable of carrying out logic resource migration and an existing cache content fragment set of the vehicle node which is updated periodically, and the virtual machine has vehicle residual computing resources with heterogeneous different sizes and trust dependence social attributes between users and the vehicle nodes; a small number of non-trusted vehicles exist, the vehicles are divided into a host vehicle and auxiliary shared vehicles sharing resources in the edge service process, communication availability ranges exist among different vehicle nodes and among vehicle nodes and task users, and the users communicate with the aerial edge nodes, the vehicle nodes and the host vehicle and the auxiliary shared vehicles through wireless communication links; the air edge node and the vehicle node also serve as edge cache nodes to provide content for users with content acquisition requirements.
Further, in step 2, the specific process of constructing the edge service knowledge graph is as follows: step 2.1, abstracting the air-ground collaboration unloading and content acquisition model network entity constructed in the step 1 into nodes and establishing different logic layers, namely a program task layer, a user layer, a vehicle node layer, an air edge node layer and a cloud node layer; step 2.2, extracting and analyzing characteristic knowledge of nodes in each layer, and extracting multi-time-varying parameter knowledge relations among the nodes in the layers; step 2.3, node attribute embedding is carried out on the intra-layer nodes, time-varying parameter knowledge relation arrangement among the nodes is carried out to construct edge relations and parameter weight factors, and multi-time-varying parameter relation arrangement is carried out on the inter-layer nodes according to different edge relation criteria to construct inter-layer node knowledge relations and parameter weight factors; step 2.4, establishing an undirected weighted graph to obtain an edge service knowledge graph; the method comprises the following steps: for multiple times by different side relation criteriaDynamic arrangement of variable parameter knowledge relationship, obtaining edge service knowledge graph, and representing the edge service knowledge graph as undirected weighted graphThe method comprises the steps of carrying out a first treatment on the surface of the Wherein (1)>Representing node set,/->Representing edges in the undirected weighted graph, +.>Representing the embedded parametric weight.
Further, in step 3, the specific procedure for establishing the system utility function weighted by the multiple performance indexes is as follows: step 3.1, constructing an air-ground communication model: the method comprises the following steps: step 3.1.1, calculating the transmission rate of the space-to-ground communication model by using a shannon formula, and respectively calculating the uplink transmission rate and the downlink transmission rate between nodes during space-to-ground communication as follows:
Wherein, during uplink communication, the nodeAs transmitting node, node->As receiving node->Representing nodesTo node->Uplink transmission rate of space-to-ground communication, +.>Representing node->To node->Bandwidth resources allocated for upstream communication between, < >>Representing node->Is set to the transmission power of (a); node ∈during downlink communication>As transmitting node, node->As receiving node->Representing node->To node->Downlink transmission rate of space-to-ground communication, +.>Representing node->To node->Bandwidth resources allocated for downstream communication between +.>Representing node->Is set to the transmission power of (a); />An additive white gaussian noise representing the time of space communication; />Representing node->And node->Path loss factor between->The calculation formula is as follows:
wherein, the liquid crystal display device comprises a liquid crystal display device,representing the carrier frequency +.>Indicating the speed of light +.>Representing node->And node->The distance between the two plates is set to be equal,representing node->And node->Probability of line-of-sight links between +.>And->Respectively representing the environmental loss of the line-of-sight link and the non-line-of-sight link; step 3.1.2, calculating the transmission rate of the ground communication model by a shannon formula; the communication transmission rate between the user and the vehicle node virtual edge is calculated as follows:
wherein, the liquid crystal display device comprises a liquid crystal display device, Representing ground emission node->And ground receiving node->The uplink transmission rate between the two,representing ground emission node->And ground receiving node->Bandwidth resources allocated between, ">Representing ground emission node->Transmit power of>Indicating channel gain, +.>Representing ground emission node->And ground receiving node->Distance between->An additive white gaussian noise representing when a user communicates with a virtual edge of a vehicle node and between vehicle nodes; step 3.2, constructing a task model; the method comprises the following steps: based on the full duplex communication technology, the user can simultaneously perform space-ground collaborative task unloading and content acquisition, and the requirements of the user are divided into calculation unloading requirements and content acquisition requirements; exist under each time slotIndividual users and each user generates only one application +.>Application->Can be divided into->A subtask with a dependency relationship, defined as an attribute tuple +.>The method comprises the steps of carrying out a first treatment on the surface of the Wherein, the liquid crystal display device comprises a liquid crystal display device,representing user +.>Type of->Representing transmission dataSize of->Representing the required calculation amount +.>Representing the maximum tolerated delay +.>Representing subtask dependency properties,>representing a subtask sequence number; in terms of user content acquisition request, there is +. >The individual demand users, the demand users are all users +.>In (2) generating a content acquisition request per user while generating a computation offload request, will +.>The individual demand user acquisition request is defined asWherein->Representing the user of the demand +.>Size of requested content->Representing the user of the demand +.>Obtaining the preference degree of the content->Cable representing content segmentsGuide (S)/(S)>Representing the user of the demand +.>Obtaining the maximum tolerable time delay of the content; step 3.3, constructing a calculation model; the method comprises the following steps: by the variable->、、/>、/>The method comprises the steps of respectively representing four task unloading modes of a user, an aerial edge node, vehicle virtual resource sharing and remote cloud, wherein each subtask can only select one unloading mode; step 3.3.1, user self unloading mode: subtask->Completion delay of offloading at user node itself +.>Expressed as:
wherein, the liquid crystal display device comprises a liquid crystal display device,representing user +.>Subtask->The amount of calculation required +.>Representing user +.>Is a computing resource of (a); meanwhile, the user uninstalls the leasing resources without leasing the computing resources, and the leasing utility of the computing resources is 0; step 3.3.2, aerial edge node unloading mode: subtask->Offloading to an over-the-air edge node->Is->Expressed as:
wherein, the liquid crystal display device comprises a liquid crystal display device, Representing the total number of edge nodes over the air, +.>Representation and air edge node->Is used to determine the correlation factor of (a),representing user +.>Subtask->Transmission data size, +.>Representing the user +.>Subtasks of->Uplink transmission to an over-the-air edge node->Transmission rate of>Representing user +.>Subtask->The amount of calculation required +.>Representing the air edge node +.>Assigned to user->Neutron task->Is proportional to the computing resource of->Representing the air edge node +.>Is a residual computing resource of (1); user->Subtask->Leasing utility generated by leasing air edge node computing resources>Expressed as:
wherein, the liquid crystal display device comprises a liquid crystal display device,representing lease price of computing resources of the aerial edge node; step 3.3.3, sharing and unloading mode of virtual resources of the vehicle: in addition to considering different idle resources of vehicles, social trust relationship between users and a vehicle-mounted self-organizing network is considered, and user nodes can sense and comprehensively infer candidate host vehicles and auxiliary shared vehicles through a knowledge graph, dynamically arrange virtual logic resources on virtual machines to construct a stable shared virtual resource pool, and the pool computing resources are shared>Expressed as:
wherein, the liquid crystal display device comprises a liquid crystal display device,representing user and mobile vehicle node->Trust between dependent on social properties->Representing a mobile vehicle node +. >Left computing resources of->Factor indicating whether it is an auxiliary shared vehicle, +.>Representing auxiliary shared vehicle->Trust dependent social attributes, +.>Representing auxiliary shared vehicle->Is a computing resource of (a); task completion delay for offloading subtasks to vehicle virtual resource sharing +.>Expressed as:
wherein, the liquid crystal display device comprises a liquid crystal display device,representing user +.>Type of->Representing the total number of nodes of the mobile vehicle, +.>Factor indicating whether or not it is a candidate host vehicle +.>Representing user +.>Subtasks of->Uplink to mobile vehicle node->Transmission rate of>Representing user +.>Subtask->Transmission data size, +.>Representing user +.>Subtask->The amount of calculation required; user->Subtask->Rental utility generated by rental vehicle virtual computing resources>Expressed as:
wherein, the liquid crystal display device comprises a liquid crystal display device,representing a vehicle node computing resource lease price; step 3.3.4, remote cloud node unloading mode: the cloud can process a plurality of subtasks simultaneously, and the completion time delay of the subtasks to be unloaded to a remote cloud node is +.>The method comprises the following steps:
wherein, the liquid crystal display device comprises a liquid crystal display device,representing air edge nodes and users->Cover factor of->Indicating the size of the data to be transmitted,representing ∈k from the air edge node>Uplink transmission rate to base station, +.>Representing user +.>Subtask- >Uplink transmission rate to base station, +.>Indicating uplink transmission rate between base station and cloud, < >>Representing cloud allocation to users->Subtask->Is proportional to the computing resource of->Representing remaining computing resources of the remote cloud node; user->SubtasksLeasing utility generated by leasing remote cloud computing resources>Expressed as:
wherein, the liquid crystal display device comprises a liquid crystal display device,representing time slot->Calculating the lease price of the resource by the remote cloud node; step 3.4, constructing a content acquisition model; the method comprises the following steps: the presence of content acquisition requirements where part of the users may have different access preferences while taking user task offloading into account, content requirement mobile users can choose to acquire preferred content from vehicle nodes, air edge nodes and remote cloud nodes, let ∈ ->、/>、/>Three different acquisition modes of a vehicle node, an air edge node and a remote cloud node are respectively represented; step 3.4.1, vehicle node acquisition mode: demand user->From the mobile vehicle node->Completion delay of acquiring content>Expressed as:
wherein, the liquid crystal display device comprises a liquid crystal display device,representing the user of the demand +.>Size of requested content->Representing the user of the demand +.>And mobile vehicle node->Downlink transmission rate between the two; rental utility of requiring a user to download vehicle node content>Expressed as:
Wherein, the liquid crystal display device comprises a liquid crystal display device,rental price indicating that the demand user obtains the desired content from the vehicle node,/->Representing the user of the demand +.>Acquiring the preference degree of the content; step 3.4.2, air edge node acquisition mode: demand user->From the air edge node->Completion delay of acquiring content>Expressed as:
wherein, the liquid crystal display device comprises a liquid crystal display device,representing the user's need for content retrieval>And air edge node->Is used to determine the correlation factor of (a),representing the air edge node +.>Is +.>Downlink transmission rate between the two; rental utility of requiring users to download over-the-air edge node content>Expressed as:
wherein, the liquid crystal display device comprises a liquid crystal display device,representing lease unit price for requiring a user to acquire content from an aerial edge node; step 3.4.3, remote cloud node acquisition mode: demand user->Completion delay of obtaining content from remote cloud node +.>Expressed as:
wherein, the liquid crystal display device comprises a liquid crystal display device,and->Representing cloud node to base station and base station to demand user +.>Is a downlink transmission rate of (a); demand user->Rental utility of downloading remote cloud node content +.>Expressed as:
wherein, the liquid crystal display device comprises a liquid crystal display device,representing time slot->Requiring the user to acquire the lease price of the content from the remote node; step 3.5, establishing a system utility function weighted by multiple performance indexes; the specific process is as follows: step 3.5.1, by taking into account the different possible offloading modes, slot +. >Delay and +.>Expressed as:
wherein, the liquid crystal display device comprises a liquid crystal display device,representing subtasks->Dependency of attribute factor->Is->Dependency variable value at attribute time, +.>Representing the total amount of subtask dependent properties +.>Representing user +.>Subtask->Selecting self-unloading->Representing subtasks->Delay in the completion of the self-offloading of the user node +.>Representing user +.>Subtask->Selecting over-the-air edge node offload,>representing subtasks->Offloading to an over-the-air edge node->Completion delay of->Representing user +.>Subtask->Selecting vehicle virtual resource sharing offload, +.>Task completion latency representing offloading of subtasks to vehicle virtual resource sharing +.>Representing user +.>Subtask->Selecting remote cloud for task offloading, +.>Representing the completion time delay of the subtask offloading to the remote cloud node; step 3.5.2, leasing computing resource leasing utility generated by edge node>Expressed as:
wherein, the liquid crystal display device comprises a liquid crystal display device,representing user +.>Subtask->Leasing the air edge node computing resource generated leasing utility,representing user +.>Subtask->Rental utility generated by rental vehicle virtual computing resources, +.>Representing a userSubtask->Leasing the leasing utility generated by the remote cloud computing resource; step 3.5.3, combining different content acquisition strategies, requiring the user to be in slot +. >Content acquisition delay and->Expressed as:
wherein, the liquid crystal display device comprises a liquid crystal display device,representing the user of the demand +.>Selecting a vehicle node content acquisition mode,/->Representing the user of the demand +.>Selecting an over-the-air edge node content acquisition mode, +.>Representing the user of the demand +.>Selecting a remote cloud node content acquisition mode; step 3.5.4, service rental utility of user-acquired edge node content associated with user demand content size and popularity ∈>Expressed as:
wherein, the liquid crystal display device comprises a liquid crystal display device,rental utility indicating that the user is required to download the contents of the vehicle node,/->Rental utility indicating that the user is required to download the content of the edge node in air,/->Representing leasing utility of a demand user to download remote cloud node content; step 3.5.5, quantitatively analyzing and calculating a system utility function represented by a weighted sum of unloading and content acquisition time delay, and the expense lease utility of user lease edge resources and content acquisition service expense lease utility->:
Wherein, the liquid crystal display device comprises a liquid crystal display device,、/>、/>、/>respectively represent time slots->Delay and +.>Computing resource lease utility generated by lease edge node>The demand user is in time slot->Content acquisition delay and->Service rental utility of user obtaining edge node content>Weight coefficient of>、/>Representing computing resource lease utility and acquisition edges, respectively The edge node content service leases a discount factor for utility.
Further, in step 4, the specific process of obtaining the optimal strategy is as follows: step 4.1, constructing an optimization problem; the method comprises the following steps: minimizing computing unloading delay, content acquisition delay of edge service, leasing edge computing resources by users and leasing edge content utility expenditure and achieving the goal of maximizing long-term network utility; definition of the definitionRepresents the maximum number of time slots,/->Representing a non-preset angle; />Representing user +.>Subtasks of->Selecting a user self, an air edge node, vehicle virtual resource sharing and remote cloud node computing unloading mode and requiring users +.>Selecting three content acquisition modes of a vehicle node, an aerial edge node and a remote cloud node; />Indicating that a user can only select one mode to perform task unloading and content acquisition; />Subtask representing user->When different computing task unloading is selected, only one air edge node and one vehicle node are selected as a host vehicle, and a user is required to be +.>When content acquisition is carried out through the aerial edge node, only one node is selected; />Indicating that the distance between the air edge nodes is not less than the minimum safe distance, and the non-preset angle change range is +. >;/>Indicating that the mobile positions of the air edge node, the vehicle node and the user do not exceed the set area limit; />A scale factor representing the allocation of computing resources; />Andindicating that the distribution of computing resources does not exceed the total computing resources of the user node, the air edge node and the remote cloud node during unloading respectively; />Indicating that the task calculation unloading time delay does not exceed the maximum unloading tolerance time delay, and the user content request time delay is smaller than or equal to the maximum unloading tolerance time delay of the content request; the original optimization problem established is expressed as +.>:
Wherein, the liquid crystal display device comprises a liquid crystal display device,representing the user's need for content retrieval>And air edge node->Is a correlation factor of (2); />Representing the air edge node +.>And another air edge node->A distance therebetween; />Representing a minimum security distance between adjacent air edge nodes; />Representing the air edge node +.>Is a horizontal axis coordinate value; />Representing user +.>Is a horizontal axis coordinate value;representing a mobile vehicle node +.>Is a horizontal axis coordinate value; />A horizontal axis coordinate value representing the coverage area boundary of the macro base station;representing the air edge node +.>Is a vertical axis coordinate value of (2); />Representing user +.>Is a vertical axis coordinate value of (2); />Representing a mobile vehicle node +.>Is a vertical axis coordinate value of (2); />Representing the vertical axis coordinate value of the coverage area boundary of the macro base station; / >Representing user +.>Subtasks of->Is used for unloading calculation time delay; />Representing user +.>Subtasks of->Unloading the maximum tolerant delay; />Representing a demand userSelecting content acquisition time delay generated by adding different content acquisition modes; />Representing the user of the demand +.>Maximum tolerant time delay of content acquisition; step 4.2, optimizing the solution: first, the original optimization problem->The medium discrete variable is relaxed to be changed into a continuous interval variable; second, introducing an upper-bound relaxation variable to the maximum nonlinear term in the objective function>Converting it into linear term while adding new constraint +.>Optimization problem after relaxation->Is->Performing an equivalent solution; the optimization problem is expressed as follows after simplification>:
Wherein, the liquid crystal display device comprises a liquid crystal display device,representing user +.>Subtasks of->Selecting leasing utilities of computing resources generated by adding different unloading modes; />Representing the user of the demand +.>Selecting content acquisition time delay generated by adding different content acquisition modes; />Representing the user of the demand +.>Selecting a content service lease utility generated by adding different content acquisitions; />Represents a relaxation of four unloading and three content acquisition discrete variables into a continuous variable between 0 and 1,/v>Representing the user +. >The dependency attribute factor in all subtasks is +.>Upper bound relaxation variable constraint of ∈10->Representing subtasks->Dependency of attribute factor->Is->Dependency variable value at attribute time, +.>Representing user +.>The dependency attribute factor in all subtasks is +.>Upper limit relaxation variable of (2);
the optimization problem is decomposed into three sub-problems by a block coordinate descent method and a successive approximation algorithm: user computing offload and content acquisition variable sub-questionsSub-problem of calculating the resource allocation proportion variable ≡>And the sub-problem of deployment angle variable of the air node->The method comprises the steps of carrying out a first treatment on the surface of the The decomposition of a specific sub-problem is represented as follows: step 4.2.1, giving a non-preset angle and calculating a resource allocation proportion strategy, and solving an unloading and content acquisition strategy; />
Wherein, the liquid crystal display device comprises a liquid crystal display device,a discount factor representing the utility of a lease of a computing resource, +.>Representing a discount factor for obtaining the lease utility of the edge node content service; step 4.2.2, given task unloading, content acquisition and non-preset angle strategies, solving a calculation resource allocation proportion strategy; because no variable coupling relation exists between the distribution of the computing resources and the acquisition of the content, the optimization problem is further simplified to obtain the sub-problem +.>Still belonging to the same solution problem;
step 4.2.3, giving task unloading, content acquisition and calculating a resource allocation proportion strategy, and solving an optimal track strategy of the aerial edge node; optimization problem with respect to post relaxation And variable influence analysis, further simplifying the problem into sub-problems with the same solution when solving the trajectory optimization strategy>;
Wherein, the liquid crystal display device comprises a liquid crystal display device,representing a non-preset angle +.>In the range of 0 to->Between (I)>Representing the air edge node +.>And another air edge node->Distance between->Not lower than the minimum safe distance; finally, by giving a part of optimized variable parameters, carrying out variable relaxation by combining Taylor expansion of local points, solving different sub-problems after converting a non-convex optimization problem into a convex optimization problem, and comparing multiple iterations with a set threshold value to obtain the optimizationTheoretical optimal boundary solution of the problem; step 4.3, optimizing and perceiving decision analysis under deployment and shared resource pooling based on continuous clipping near-end strategies: first, the initial network state of the system is defined as +.>:/>
Wherein, the liquid crystal display device comprises a liquid crystal display device,representing user +.>Is (are) moved positions>Representing user +.>Is>Representing user +.>Transition probability of->Representing user +.>Is->Representing the user of the demand +.>Is (are) acquired request>Representing user +.>Task model with task topology knowledge relationship, +.>Representing a mobile vehicle node +.>Is (are) located>Representing a mobile vehicle node +.>Is>Representing a mobile vehicle node +. >Trust dependent social attributes, +.>Representing a mobile vehicle node +.>Left computing resources of->Representing a mobile vehicle node +.>Existing cached content piece set,/->Represent the firstPost-deployment location of individual air edge nodes, +.>Representing the speed of movement of the edge node in air, +.>Representing the air edge node +.>Left computing resources of->Representing the air edge node +.>Part of the existing cached content set,/->Indicating macro base station coordinate position,/->Representing remaining computing resources of remote cloud node, +.>Representing a cached content segment set in the cloud; analyzing a plurality of knowledge factors influencing the quality of service performance, arranging the dynamic knowledge relationship to construct an edge service knowledge graph, and obtaining preprocessing knowledge state information expressed as +.>The method comprises the steps of carrying out a first treatment on the surface of the Second, define the complex motion space under each time slot as +.>Wherein, the method comprises the steps of, wherein,representing user +.>Subtasks of->Different offloading policies of->Representing different content acquisition strategies of a demand user; the set bonus function under each time slot is denoted +.>。
Further, in step 4.3, the overall flow of space-to-ground collaborative offloading and content acquisition under perceived deployment and shared resource pooling optimized based on continuous clipping near-end policy is as follows: step 4.3.1, constructing an air-ground collaboration unloading and content acquisition model composed of remote cloud, non-preset track air edge nodes, road mobile vehicles and two types of mobile users with different types and different requirements, and initializing parameters; step 4.3.2, executing a training round, and initializing a training model to obtain an initial state; step 4.3.3, executing time slot rounds, analyzing a plurality of time-varying parameter relation arrangement of nodes in layers and among layers to construct an edge service knowledge graph, and obtaining knowledge state information; step 4.3.4, the knowledge server selects an action strategy through a strategy network; step 4.3.5, the action strategy is put into the environment for execution, and rewards under the current network state, the next network state, the system utility function, the unloading and content acquisition strategy and the stored experience tuples are obtained; step 4.3.6, judging whether the parameters of the current strategy network need to be updated, if so, entering another strategy network and evaluating the network to carry out training update, otherwise, continuously executing and updating the network state; 4.3.7, if all time slot training is finished, calculating average network rewards, finishing one training round, and initializing a training model; and 4.3.8, if the training round is finished, obtaining an average network reward and an optimal service strategy, and outputting the space cooperation unloading and content acquisition scheme as an optimal scheme.
The beneficial technical effects brought by the invention are as follows: according to the invention, a 5G macro base station is considered to cover a remote cloud, an aerial edge node, a mobile vehicle and a vehicle-mounted self-organizing network edge urban space-ground cooperation environment with concurrent different types of user demands, an edge service knowledge graph is dynamically arranged and constructed based on edge node attributes and a plurality of time-varying parameter knowledge relations among the nodes, a system utility function weighted by multiple performance indexes in four edge unloading modes and three content acquisition modes is established according to edge node perception deployment and a vehicle-mounted self-organizing network virtual resource sharing pooling knowledge model, and a space-ground cooperation unloading and content acquisition strategy based on edge service knowledge graph perception is constructed for the different types of user demands by using a theoretical optimization and cutting near-end strategy optimization algorithm. The invention solves the problems of space-to-ground collaborative unloading and content acquisition of different types of users in the vehicle-mounted self-organizing network edge city environment by taking into consideration the dynamic arrangement of a plurality of time-varying parameter knowledge between the vehicle edge environment equipment, the perceived deployment of the aerial edge nodes and the method for establishing the vehicle virtual shared resource pool in the edge nodes and the users from the angles of the dependency task unloading requirements and the different content acquisition requirements of different types of users, thereby effectively reducing the edge service delay and the lease utility and improving the service quality requirements of the edge users while maximally arranging and utilizing the edge idle resources.
Drawings
Fig. 1 is a schematic block diagram of a space-ground collaboration unloading and content acquisition method based on knowledge-graph perception.
Fig. 2 is a schematic diagram of a space-ground collaboration unloading and content acquisition model based on knowledge-graph perception.
FIG. 3 is a schematic diagram of an edge service knowledge aware layer model according to the present invention.
Detailed Description
The invention is described in further detail below with reference to the attached drawings and detailed description:
as shown in fig. 1, a space-ground collaboration unloading and content obtaining method based on knowledge-graph perception comprises the following steps:
step 1, constructing an edge vehicle-mounted self-organizing network environment hollow cooperative unloading and content acquisition model consisting of a macro base station, a remote cloud, a non-preset track aerial edge node, a road mobile vehicle and two types of mobile users with different types and different requirements in a 5G macro base station coverage lower edge urban environment. The invention considers the edge city of the 5G macro base station covering remote cloud, air edge node, road mobile vehicle and two kinds of mobile users with different types and different demandsThe space-city collaboration scenario is shown in fig. 2. The macro base station is provided with an edge knowledge server for knowledge spectrum sensing, can sense physical entity information of users, vehicles and aerial edge nodes in a coverage area, is connected with a remote cloud in an optical fiber communication mode, and contains a large amount of computing resources and cache content fragment sets In the present embodiment,/>Indicate->Content piece->Representing the total number of content segments. Meanwhile, quantitatively marking the positions of different physical entities in the scene by establishing a three-dimensional coordinate system, and expressing the coordinate positions of the macro base station as。/>Unmanned aerial vehicle of a fixed height (+)>The unmanned plane is->An air edge node, thus use->Representing the total number of edge nodes in air) at a non-preset angle +.>Flexible perceptive decision deployment of flights as airborne edge nodesFirst->The position of each air edge node after deployment is +.>The minimum security distance between adjacent air edge nodes is +.>Partial existing cache content set containing computing resources and periodic updates in air edge node +.>To assist the requesting user in task offloading, relay forwarding and providing cached content, in this embodiment +.>The air edge node communicates with the macro base station via an air link. Two kinds of task users with different moving speeds and tolerant time delays exist in the scene, namely a common user and a vehicle-mounted user, and the users are +.>The position is expressed as +.>. Co-existence of +.>Vehicle moving vehicle, moving vehicle node +.>The position of (2) is expressed as +.>Mobile vehicle node +. >Is expressed as +.>Each vehicle carries a virtual machine capable of logical resource migration and a periodically updated mobile vehicle node +.>Existing cached content fragment set->Mobile vehicle nodes with heterogeneous and different sizes in virtual machine +.>Is->User and mobile vehicle node->Trust-dependent social properties between->There are a small number of non-trusted vehicles and auxiliary shared vehicles in which the vehicles are to be divided into a host vehicle and shared resources during edge service, there is a range of availability of communication between different vehicle nodes, between vehicle nodes and task users, and communication between the users and air edge nodes, between the vehicle nodes and the host vehicle and auxiliary shared vehicles is performed by wireless communication links. The air edge node and the vehicle node also serve as edge cache nodes to provide content for users with content acquisition requirements.
In the invention, the weighting of the task calculation unloading time delay, the content acquisition time delay, the lease computing resource expense utility and the lease edge content expense utility is used for representing the system utility function. The invention has the core problems of how to synthesize the multi-time-varying parameter dynamic knowledge relationship among the edge devices to construct the edge service knowledge graph and acquire the optimal edge service strategy of the demand user so as to effectively reduce the edge service time delay and lease utility expenditure and improve the service quality demand of the edge user.
And 2, dynamically arranging and constructing an edge service knowledge graph based on the edge node attribute and a plurality of time-varying parameter knowledge relations among the nodes. An edge service knowledge perception layer model schematic diagram established by abstracting and synthesizing a plurality of time-varying parameter logic knowledge is shown in fig. 3, wherein an edge logic layer in the knowledge perception layer model comprises a program task layer, a user layer, a vehicle node layer, an air edge node layer and a cloud node layer. In the air edge node layer, each air edge node analyzes a distance parameter relation with adjacent air nodes to establish an intra-layer knowledge link for information transfer; in the user layer, an intra-layer knowledge link is established between the vehicle-mounted user node and the common user node by analyzing the inter-node distance parameter relation to carry out information transfer; in a vehicle node layer, establishing an intra-layer knowledge link between a main vehicle node, an untrusted vehicle node and an auxiliary shared vehicle node by analyzing a social trust relationship, a relative moving speed, connection availability and other time-varying parameter relationships to carry out information transfer, so that vehicle virtual resource sharing is formed, and the intra-layer knowledge link between the untrusted vehicle node and the auxiliary shared vehicle node which is unavailable in connection availability cannot be established; analyzing the dependency relationship between the user node application subtasks and different subtasks in the program task layer to establish an intra-layer knowledge link; and performing task mapping between the user layer and the program task layer, and performing virtual resource migration mapping between the vehicle node layer and the virtual resource pool. The user layer can establish an uplink unloading interlayer logic selection strategy, and the user layer, the cloud node layer, the air edge layer and the vehicle node layer can respectively establish the uplink unloading interlayer logic selection strategy and the downlink content acquisition interlayer logic selection strategy. The knowledge server can collect the perception information of different edge logic layers and perceive the related information such as edge nodes, users, idle resources, application program task files and the like in the coverage area. The specific process is as follows:
Step 2.1, abstracting the air-ground collaboration unloading and content acquisition model network entity constructed in the step 1 into nodes and establishing different logic layers, namely abstracting different network entities in an edge city environment to obtain different logic layers, namely a program task layer, a user layer, a vehicle node layer, an air edge node layer and a cloud node layer;
step 2.2, extracting and analyzing node characteristic knowledge in each layer, and simultaneously extracting multi-time-varying parameter knowledge relations among the layer nodes, namely extracting user abstract node positions, moving speeds, transition probabilities, program task mapping relations and content request knowledge characteristics, and simultaneously extracting edge attribute knowledge characteristics such as content caching, idle resources, moving speeds, connection availability, social attribute trust relations and the like of the edge nodes;
step 2.3, node attribute embedding is carried out on the intra-layer nodes, time-varying parameter knowledge relation arrangement among the nodes is carried out to construct edge relations and parameter weight factors, and multi-time-varying parameter relation arrangement is carried out on the inter-layer nodes according to different edge relation criteria to construct inter-layer node knowledge relations and parameter weight factors;
step 2.4, establishing an undirected weighted graph to obtain an edge service knowledge graph; the method comprises the following steps: the edge service knowledge graph is obtained through dynamic arrangement of different side relation criteria on multi-time-varying parameter knowledge relations and is expressed as an undirected weighted graph . Wherein (1)>Representing node set,/->Representing edges in the undirected weighted graph, +.>Representing the embedded parametric weight. The same different nodes can cache the content required by the user, and the user content acquisition requirement can be better served through the establishment of time-varying parameter logic knowledge.
And step 3, analyzing the influence factors of the user service quality under different service strategies and establishing a multi-performance index weighted system utility function. The specific process is as follows:
step 3.1, constructing an air-ground communication model; the method comprises the following steps: step 3.1.1, calculating the transmission rate of the space-to-ground communication model by using a shannon formula, and respectively calculating the uplink transmission rate and the downlink transmission rate between nodes during space-to-ground communication as follows:
wherein, during uplink communication, the nodeAs transmitting node, node->As receiving node->Representing nodesTo node->Uplink transmission rate of space-to-ground communication, +.>Representing node->To node->Bandwidth resources allocated for upstream communication between, < >>Representing node->Is set to the transmission power of (a); node ∈during downlink communication>As transmitting node, node->As receiving node->Representing node->To node->Downlink transmission rate of space-to-ground communication, +.>Representing node->To node- >Bandwidth resources allocated for downstream communication between +.>Representing node->Is set to the transmission power of (a); />An additive white gaussian noise representing the time of space communication; />Representing node->And node->Path loss factor between->The calculation formula is as follows:
wherein, the liquid crystal display device comprises a liquid crystal display device,representing the carrier frequency +.>Indicating the speed of light +.>Representing node->And node->The distance between the two plates is set to be equal,representing node->And node->Probability of line-of-sight links between +.>And->The environmental loss of the line-of-sight link and the non-line-of-sight link are represented, respectively.
Step 3.1.2, calculating the transmission rate of the ground communication model by a shannon formula; the communication transmission rate between the user and the vehicle node virtual edge communication and the vehicle node can be calculated as:
wherein, the liquid crystal display device comprises a liquid crystal display device,representing ground emission node->And ground receiving node->The uplink transmission rate between the two,representing ground emission node->And ground receiving node->Bandwidth resources allocated between, ">Representing ground emission node->Transmit power of>Indicating channel gain, +.>Representing ground emission node->And ground receiving node->Distance between->Representing additive white gaussian noise when a user communicates with a virtual edge of a vehicle node and between vehicle nodes.
Step 3.2, constructing a task model; the method comprises the following steps: based on full duplex communication technology, users can simultaneously perform space-ground collaborative task unloading and content acquisition, and the method is to be used The demands of the user are divided into computational offload demands and content acquisition demands. Exist under each time slotIndividual users and each user generates only one application +.>Application->Can be divided into->A subtask with a dependency relationship, defined as an attribute tuple +.>. Wherein, the liquid crystal display device comprises a liquid crystal display device,representing user +.>Type of->Representing the size of the transmitted data +.>Representing the required calculation amount +.>Representing the maximum tolerated delay +.>Representing subtask dependency properties,>representing the subtask sequence number. In terms of user content acquisition request, there is +.>Individual demand users (demandThe user is all users->Some of the users in (a), each user generates a content acquisition request while generating a computation offload request, will +.>The individual demand user acquisition request is defined asWherein->Representing the user of the demand +.>Size of requested content->Representing the user of the demand +.>Obtaining the preference degree of the content->Index representing content piece (i.e.)>Content piece), ->Representing the user of the demand +.>The maximum tolerable delay of the content is obtained.
Step 3.3, constructing a calculation model; the method comprises the following steps: by variable amounts、/>、/>、The method respectively represents four task unloading modes of a user, an aerial edge node, vehicle virtual resource sharing and remote cloud, and each subtask can only select one unloading mode.
Step 3.3.1, user self unloading mode: order theRepresenting user +.>Is a subtask->Completion delay of offloading at user node itself +.>Can be expressed as:
wherein, the liquid crystal display device comprises a liquid crystal display device,representing user +.>Subtask->The amount of computation required. Meanwhile, the user uninstalls the computing resource without leasing the computing resource, and the leasing utility of the computing resource is 0.
Step 3.3.2, aerial edge node unloading mode: order theRepresenting the total number of edge nodes over the air, +.>Representation and air edge node->Is associated with (a) factor(s)>Representing the air edge node +.>Assigned to user->Neutron task->Is proportional to the computing resource of->Representing the air edge node +.>Left computing resources of->Representing the lease price for the computing resource of the edge node in air. Subtask->Offloading to an over-the-air edge node->Is->Can be expressed as:
wherein, the liquid crystal display device comprises a liquid crystal display device,representing user +.>Subtask->Transmission data size, +.>Representing the user +.>Is of the sub-task of (a)Uplink transmission to an over-the-air edge node->Transmission rate of>Representing user +.>Subtask->The amount of computation required. User->Subtask->Leasing utility generated by leasing air edge node computing resources>Expressed as:
step 3.3.3, sharing and unloading mode of virtual resources of the vehicle: in addition to considering different idle resources of the vehicle, the user and the vehicle-mounted self-organization are also considered Social trust relationship between fabric networks, letRepresenting user and mobile vehicle node->Trust between dependent on social properties->Representing a mobile vehicle node +.>Left computing resources of->Factor indicating whether or not it is a candidate host vehicle +.>Factor indicating whether it is an auxiliary shared vehicle, +.>Representing a vehicle node computing resource lease price. The user node can perceive and comprehensively infer candidate host vehicles and auxiliary shared vehicles through the knowledge graph, dynamically arrange virtual logic resources on the virtual machine to construct a stable shared virtual resource pool, and pool computing resources +.>Can be expressed as:
wherein, the liquid crystal display device comprises a liquid crystal display device,representing auxiliary shared vehicle->Trust dependent social attributes, +.>Representing auxiliary shared vehiclesIs a computing resource of (a). Task completion delay for offloading subtasks to vehicle virtual resource sharing +.>Expressed as:
wherein, the liquid crystal display device comprises a liquid crystal display device,representing user +.>Subtasks of->Uplink to mobile vehicle node->Is used for the transmission rate of (a),representing user +.>Subtask->Transmission data size, +.>Representing user +.>Type of->Representing user +.>Subtask->The amount of computation required. User->Subtask->Rental utility generated by rental vehicle virtual computing resources>Expressed as:
step 3.3.4, remote cloud node unloading mode: the cloud can process a plurality of subtasks simultaneously to enable Representing air edge nodes and users->Cover factor of->Representing cloud allocation to users->Subtask->Is proportional to the computing resource of->Representing remaining computing resources of remote cloud node, +.>Representing time slot->The remote cloud node calculates a lease price for the resource. Completion delay of subtask offloading to remote cloud node +.>Expressed as:
wherein, the liquid crystal display device comprises a liquid crystal display device,representing ∈k from the air edge node>Uplink transmission rate to base station, +.>Representing user +.>Subtask->Uplink transmission rate to base station, +.>Representing the uplink transmission rate between the base station and the cloud. User->SubtasksLeasing utility generated by leasing remote cloud computing resources>Expressed as:
step 3.4, construction Contents acquisitionTaking a model; the method comprises the following steps: content acquisition requirements where there may be different access preferences for some users while considering user task offloading, content requirement mobile users can choose to acquire preferred content from vehicle nodes, air edge nodes, and remote cloud nodes, to have、/>、/>Three different acquisition modes are represented, respectively, for the vehicle node, the air edge node, and the remote cloud node.
Step 3.4.1, vehicle node acquisition mode: order theRepresenting a mobile vehicle node +.>Periodically updated existing cached content piece set,/-for >Rental price indicating that the demand user obtains the desired content from the vehicle node,/->Representing the user of the demand +.>And mobile vehicle node->Downlink transmission rate between the two. Demand user->From the mobile vehicle node->Completion delay of acquiring content>Expressed as:
wherein, the liquid crystal display device comprises a liquid crystal display device,representing the user of the demand +.>The size of the requested content. Rental utility of requiring a user to download vehicle node content>Expressed as:
wherein, the liquid crystal display device comprises a liquid crystal display device,representing the user of the demand +.>The preference degree of the content is acquired.
Step 3.4.2, air edge node acquisition mode: order theRepresenting the air edge node +.>Part of the existing cached content set,/->Representing the user's need for content retrieval>And air edge node->Is associated with (a) factor(s)>Lease price indicating that the user is required to acquire content from an over-the-air edge node,/->Representing the air edge node +.>Is +.>Downlink transmission rate between the two. Demand user->From the air edge node->Completion delay of acquiring content>Can be expressed as:
step 3.4.3, remote cloud node acquisition mode: order theRepresenting time slot->Requiring the user to obtain the rental price of the content from the remote node, < +.>And->Representing cloud node to base station and base station to demand user +. >Is used for the downlink transmission rate of (a). Demand user->Completion delay of obtaining content from remote cloud node +.>Can be expressed as:
step 3.5, establishing a system utility function weighted by multiple performance indexes; the specific process is as follows:
step 3.5.1, time slots are allocated by taking into account the different possible offloading modesDelay and +.>Expressed as:
wherein, the liquid crystal display device comprises a liquid crystal display device,representing subtasks->Dependency of attribute factor->Is->Dependency variable value at attribute time, +.>Representing the total amount of subtask dependent properties +.>Representing user +.>Subtask->Selecting self-unloading->Representing subtasks->Delay in the completion of the self-offloading of the user node +.>Representing user +.>Subtask->Selecting an over-the-air edge nodeThe unloading is carried out by the device,representing subtasks->Offloading to an over-the-air edge node->Completion delay of->Representing user +.>Subtask->Selecting vehicle virtual resource sharing offload, +.>Task completion latency representing offloading of subtasks to vehicle virtual resource sharing +.>Representing user +.>Subtask->Selecting remote cloud for task offloading, +.>Representing the completion latency of the subtask offloading to the remote cloud node.
wherein, the liquid crystal display device comprises a liquid crystal display device,representing user +.>Subtask->Leasing the air edge node computing resource generated leasing utility,representing user +.>Subtask->Rental utility generated by rental vehicle virtual computing resources, +.>Representing user +.>Subtask->Leasing a leasing utility generated by a remote cloud computing resource.
Step 3.5.3, combining different content acquisition strategies, requiring the user to be in the time slotContent acquisition delay and->The method comprises the following steps:
wherein, the liquid crystal display device comprises a liquid crystal display device,representing the user of the demand +.>Selecting a vehicle node content acquisition mode,/->Representing the user of the demand +.>Selecting an over-the-air edge node content acquisition mode, +.>Representing the user of the demand +.>A remote cloud node content acquisition mode is selected.
Step 3.5.4, obtaining service lease utility of edge node content by user associated with user demand content size and popularityExpressed as:
wherein, the liquid crystal display device comprises a liquid crystal display device,rental utility indicating that the user is required to download the contents of the vehicle node,/->Rental utility indicating that the user is required to download the content of the edge node in air,/->Representing rental utility for requiring the user to download the remote cloud node content.
Step 3.5.5, quantitatively analyzing and calculating a system utility function represented by a weighted sum of unloading and content acquisition time delay, expenditure costs of user leasing edge resources and expenditure costs of content acquisition services :
Wherein, the liquid crystal display device comprises a liquid crystal display device,、/>、/>、/>respectively represent time slots->Delay and +.>Computing resource lease utility generated by lease edge node>The demand user is in time slot->Content acquisition delay and->Service rental utility of user obtaining edge node content>Weight coefficient of>、/>Representing computing resource lease utility and acquiring edge nodes, respectivelyA discount factor for rental utility of content services.
And 4, constructing an air-ground collaborative unloading and content acquisition optimal strategy based on edge service knowledge graph perception for different types of user demands by adopting a theoretical optimization and clipping near-end strategy optimization algorithm.
Step 4.1, constructing an optimization problem; the method comprises the following steps: minimizing computing offload latency for edge services, content acquisition latency, user leased edge computing resources and leased edge content utility expenditures and achieving the goal of long-term network utility maximization. The original optimization problem established is expressed as:/>
Wherein, the liquid crystal display device comprises a liquid crystal display device,represents the maximum number of time slots,/->Representing the user's need for content retrieval>And air edge node->Is a correlation factor of (2); />Representing user +.>Subtasks of->Selecting a user self, an air edge node, vehicle virtual resource sharing and remote cloud node computing unloading mode and requiring users +. >Selecting three content acquisition modes of a vehicle node, an aerial edge node and a remote cloud node; />Indicating that a user can only select one mode to perform task unloading and content acquisition; />Subtask representing user->When different computing task unloading is selected, only one air edge node and one vehicle node are selected as a host vehicle, and a user is required to be +.>When content acquisition is carried out through the aerial edge node, only one node is selected; />Indicating that the distance between the air edge nodes is not less than the minimum safe distance, and the non-preset angle change range is +.>;/>Indicating that the mobile positions of the air edge node, the vehicle node and the user do not exceed the set area limit; />A scale factor representing the allocation of computing resources; />Andindicating that the distribution of computing resources does not exceed the total computing resources of the user node, the air edge node and the remote cloud node during unloading respectively; />Indicating that the task calculation unloading time delay does not exceed the maximum unloading tolerance time delay, and the user content request time delay is smaller than or equal to the maximum unloading tolerance time delay of the content request; />Representing the air edge node +.>And another air edge node->A distance therebetween; />Representing a minimum security distance between adjacent air edge nodes; / >Representing the air edge node +.>Is a horizontal axis coordinate value; />Representing user +.>Is a horizontal axis coordinate value; />Representing a mobile vehicle node +.>Is a horizontal axis coordinate value; />A horizontal axis coordinate value representing the coverage area boundary of the macro base station; />Representing the air edge node +.>Is a vertical axis coordinate value of (2); />Representing user +.>Is a vertical axis coordinate value of (2); />Representing a mobile vehicle node +.>Is a vertical axis coordinate value of (2); />Representing the vertical axis coordinate value of the coverage area boundary of the macro base station; />Representing user +.>Subtasks of->Is used for unloading calculation time delay; />Representing user +.>Subtasks of->Unloading the maximum tolerant delay; />Representing the user of the demand +.>Selecting content acquisition time delay generated by adding different content acquisition modes; />Representing the user of the demand +.>The maximum tolerated delay for content retrieval of (c).
Step 4.2, optimizing the solution: first, the original optimization problemThe medium discrete variable relaxes to become a continuous interval variable. Secondly, introducing an upper-limit relaxation variable +.>Which is converted into a linear term. Definitions->Representing user +.>Subtasks of->Selecting leasing utilities of computing resources generated by adding different unloading modes; />Representing the user of the demand +.>Selecting content acquisition time delay generated by adding different content acquisition modes; / >Representing the user of the demand +.>The content service rental utilities resulting from the addition of the different content acquisitions are selected. The optimization problem is expressed as follows after simplification>:
Wherein, the liquid crystal display device comprises a liquid crystal display device,represents a relaxation of four unloading and three content acquisition discrete variables into a continuous variable between 0 and 1,/v>Representing the user +.>The dependency attribute factor in all subtasks is +.>Upper bound relaxation variable constraint of ∈10->Representing subtasks->Dependency of attribute factor->Is->The dependency variable value at the time of the attribute,representing user +.>The dependency attribute factor in all subtasks is +.>Upper bound relaxation variable of (2).
Decomposing the optimization problem into three sub-problems, namely the sub-problems of user calculation unloading and content acquisition variables, by a block coordinate descent method and a successive approximation algorithmSub-problem of calculating the resource allocation proportion variable ≡>And the sub-problem of deployment angle variable of the air node->. The decomposition of a specific sub-problem is represented as follows: />
And 4.2.1, giving a non-preset angle and calculating a resource allocation proportion strategy, and solving an unloading and content acquisition strategy.
Wherein, the liquid crystal display device comprises a liquid crystal display device,a discount factor representing the utility of a lease of a computing resource, +.>Representing a discount factor for obtaining the rental utility of the edge node content service.
And 4.2.2, giving task unloading, content acquisition and a non-preset angle strategy, and solving a computing resource allocation proportion strategy. Because no variable coupling relation exists between the distribution of the computing resources and the acquisition of the content, the optimization problem is further simplified to obtain the sub-problemStill belongs to the problem of co-solution.
And 4.2.3, giving task unloading, content acquisition and calculating a resource allocation proportion strategy, and solving an optimal track strategy of the air edge node. Optimization problem with respect to post relaxationAnd variable influence analysis, further simplifying the problem into sub-problems with the same solution when solving the trajectory optimization strategy>。
Wherein, the liquid crystal display device comprises a liquid crystal display device,representing a non-preset angle +.>In the range of 0 to->Between (I)>Representing the air edge node +.>And another air edge node->Distance between->Not lower than the minimum safe distance.
Finally, a part of optimized variable parameters are given, variable relaxation is carried out by combining Taylor expansion of local points, the non-convex optimization problem is converted into a convex optimization problem, then different sub-problems are solved, and the theoretical optimal boundary solution of the optimization problem is obtained through multiple iteration and set threshold comparison.
Step 4.3, optimizing and perceiving decision analysis under deployment and shared resource pooling based on continuous clipping near-end strategies: first, define Representing macrosBase station coordinate position->Representing remaining computing resources of remote cloud node, +.>Representing a set of cached content pieces in a cloud, defining an initial network state of the system as +.>:
Wherein, the liquid crystal display device comprises a liquid crystal display device,representing user +.>Is (are) moved positions>Representing user +.>Is>Representing user +.>Transition probability of->Representing user +.>Is->Representing the user of the demand +.>Is (are) acquired request>Representing user +.>Task model with task topology knowledge relationship, +.>Representing a mobile vehicle node +.>Is (are) located>Representing a mobile vehicle node +.>Is>Representing a mobile vehicle node +.>Trust dependent social attributes, +.>Representing a mobile vehicle node +.>Left computing resources of->Representing a mobile vehicle node +.>Existing cached content piece set,/->Indicate->In the airLocation after edge node deployment, +.>Representing the speed of movement of the edge node in air, +.>Representing the air edge node +.>Left computing resources of->Representing the air edge node +.>Is provided for the portion of the existing cached content set. Analyzing a plurality of knowledge factors influencing the quality of service performance, arranging the dynamic knowledge relationship to construct an edge service knowledge graph, and obtaining 'preprocessing' knowledge state information expressed as +. >The method comprises the steps of carrying out a first treatment on the surface of the Second, the complex motion space at each time slot is defined as +.>Wherein->Representing user +.>Subtasks of->Different offloading policies of->Representing different content acquisition policies for the requesting user. The set bonus function under each time slot is denoted +.>。
The overall flow of space-to-ground collaborative offloading and content acquisition under perceived deployment and shared resource pooling based on continuous clipping near-end policy optimization is as follows:
step 4.3.1, constructing an air-ground collaboration unloading and content acquisition model composed of remote cloud, non-preset track air edge nodes, road mobile vehicles and two types of mobile users with different types and different requirements, and initializing parameters;
step 4.3.2, executing a training round, and initializing a training model to obtain an initial state;
step 4.3.3, executing time slot rounds, analyzing a plurality of time-varying parameter relation arrangement of nodes in layers and among layers to construct an edge service knowledge graph, and obtaining knowledge state information;
step 4.3.4, the knowledge server selects an action strategy through a strategy network;
step 4.3.5, the action strategy is put into the environment for execution, and rewards under the current network state, the next network state, the system utility function, the unloading and content acquisition strategy and the stored experience tuples are obtained;
Step 4.3.6, judging whether the parameters of the current strategy network need to be updated, if so, entering another strategy network and evaluating the network to carry out training update, otherwise, continuously executing and updating the network state;
4.3.7, if all time slot training is finished, calculating average network rewards, finishing one training round, and initializing a training model;
and 4.3.8, if the training round is finished, obtaining an average network reward and an optimal service strategy, and outputting the space cooperation unloading and content acquisition scheme as an optimal scheme.
The decision method pseudocode under the perceived deployment and shared resource pooling based on continuous clipping near-end policy optimization is as follows:
1: initializing environmental parameters, experience storage pools and network rewards;
2: initializing parameters in a neural network;
9: storing the experience tuples to an experience buffer pool;
11: continuously cutting a near-end strategy to optimize, learn and update network parameters;
12: assigning the learned updated parameters to the policy network;
13: updating the network state;
14: End for;
15: calculating an average network prize;
16:End for ;
17: until the preset iteration times are executed and reach convergence;
it should be understood that the above description is not intended to limit the invention to the particular embodiments disclosed, but to limit the invention to the particular embodiments disclosed, and that the invention is not limited to the particular embodiments disclosed, but is intended to cover modifications, adaptations, additions and alternatives falling within the spirit and scope of the invention.
Claims (5)
1. The space-ground collaboration unloading and content obtaining method based on knowledge graph perception is characterized by comprising the following steps of:
step 1, constructing an edge vehicle-mounted self-organizing network environment hollow cooperative unloading and content acquisition model consisting of a macro base station, remote cloud, non-preset track aerial edge nodes, road mobile vehicles and two types of mobile users with different types and different requirements in a 5G macro base station coverage lower edge urban environment;
step 2, dynamically arranging and constructing an edge service knowledge graph based on edge node attributes and a plurality of time-varying parameter knowledge relations among nodes;
Step 3, analyzing user service quality influence factors under different service strategies and establishing a multi-performance index weighted system utility function according to node perception deployment and a vehicle-mounted self-organizing network shared resource pooling knowledge model;
in the step 3, the specific process of establishing the system utility function weighted by the multiple performance indexes is as follows:
step 3.1, constructing an air-ground communication model: the method comprises the following steps:
step 3.1.1, calculating the transmission rate of the space-to-ground communication model by using a shannon formula, and respectively calculating the uplink transmission rate and the downlink transmission rate between nodes during space-to-ground communication as follows:
wherein, during uplink communication, the nodeAs transmitting node, node->As receiving node->Representing node->To node->Uplink transmission rate of space-to-ground communication, +.>Representing node->To node->Bandwidth resources allocated for upstream communication between the two,representing node->Is set to the transmission power of (a); node ∈during downlink communication>As transmitting node, node->As a receiving node,representing node->To node->Downlink transmission rate of space-to-ground communication, +.>Representing node->To node->Bandwidth resources allocated for downstream communication between +.>Representing node->Is set to the transmission power of (a); />Representing the open spaceAdditive white gaussian noise during inter-communication; / >Representing node->And node->Path loss factor between->The calculation formula is as follows:
wherein, the liquid crystal display device comprises a liquid crystal display device,representing the carrier frequency +.>Indicating the speed of light +.>Representing node->And node->Distance between->Representing node->And node->Probability of line-of-sight links between +.>And->Respectively representing the environmental loss of the line-of-sight link and the non-line-of-sight link;
step 3.1.2, calculating the transmission rate of the ground communication model by a shannon formula;
the communication transmission rate between the user and the vehicle node virtual edge is calculated as follows:
wherein, the liquid crystal display device comprises a liquid crystal display device,representing ground emission node->And ground receiving node->Uplink transmission rate between +.>Representing ground emission node->And ground receiving node->Bandwidth resources allocated between, ">Representing ground emission node->Transmit power of>Indicating channel gain, +.>Representing ground emission node->And ground receiving node->Distance between->An additive white gaussian noise representing when a user communicates with a virtual edge of a vehicle node and between vehicle nodes;
step 3.2, constructing a task model; the method comprises the following steps:
based on the full duplex communication technology, the user can simultaneously perform space-ground collaborative task unloading and content acquisition, and the requirements of the user are divided into calculation unloading requirements and content acquisition requirements; exist under each time slot Individual users and each user generates only one application +.>Application->Can be divided into->A subtask with a dependency relationship, defined as an attribute tuple +.>The method comprises the steps of carrying out a first treatment on the surface of the Wherein (1)>Representing user +.>Type of->Representing the size of the transmitted data +.>Representing the required calculation amount +.>Representing the maximum tolerated delay +.>Representing subtask dependency properties,>representing a subtask sequence number;
random presence in ambient users in terms of user content acquisition requestsThe demand users are all usersIn (2) generating a content acquisition request per user while generating a computation offload request, will +.>The individual demand user acquisition request is defined as +.>Wherein->Representing the user of the demand +.>Size of requested content->Representing the user of the demand +.>Obtaining the preference degree of the content->Index representing content piece->Representing the user of the demand +.>Obtaining the maximum tolerable time delay of the content;
step 3.3, constructing a calculation model; the method comprises the following steps:
by variable amounts、/>、/>、/>The method comprises the steps of respectively representing four task unloading modes of a user, an aerial edge node, vehicle virtual resource sharing and remote cloud, wherein each subtask can only select one unloading mode;
step 3.3.1, user self unloading mode:
wherein, the liquid crystal display device comprises a liquid crystal display device,representing user +.>Subtask->The amount of calculation required +.>Representing user +.>Is a computing resource of (a); meanwhile, the user uninstalls the leasing resources without leasing the computing resources, and the leasing utility of the computing resources is 0;
step 3.3.2, aerial edge node unloading mode:
wherein, the liquid crystal display device comprises a liquid crystal display device,representing the total number of edge nodes over the air, +.>Representation and air edge node->Is associated with (a) factor(s)>Representing user +.>Subtask->Transmission data size, +.>Representing the user +.>Subtasks of->Uplink transmission to an over-the-air edge node->Transmission rate of>Representing user +.>Subtask->The amount of calculation required +.>Representing the air edge node +.>Assigned to user->Neutron task->Is proportional to the computing resource of->Representing the air edge node +.>Is a residual computing resource of (1); user->Subtask->Leasing utility generated by leasing air edge node computing resources>Expressed as:
wherein, the liquid crystal display device comprises a liquid crystal display device,representing lease price of computing resources of the aerial edge node;
step 3.3.3, sharing and unloading mode of virtual resources of the vehicle:
in addition to taking into account the different free resources of the vehicle, The social trust relationship between the user and the vehicle-mounted self-organizing network is considered, the user node can perceive and comprehensively infer candidate host vehicles and auxiliary shared vehicles through the knowledge graph, and dynamically arrange virtual logic resources on the virtual machine to construct a stable shared virtual resource pool, and the stable shared virtual resource pool is used for pooling computing resourcesExpressed as:
wherein, the liquid crystal display device comprises a liquid crystal display device,representing user and mobile vehicle node->Trust between dependent on social properties->Representing a mobile vehicle nodeLeft computing resources of->Factor indicating whether it is an auxiliary shared vehicle, +.>Representing auxiliary shared vehiclesTrust dependent social attributes, +.>Representing auxiliary shared vehicle->Is a computing resource of (a); task completion delay for offloading subtasks to vehicle virtual resource sharing +.>Expressed as:
wherein, the liquid crystal display device comprises a liquid crystal display device,representing user +.>Type of->Representing the total number of nodes of the mobile vehicle, +.>Factor indicating whether or not it is a candidate host vehicle +.>Representing user +.>Subtasks of->Uplink to mobile vehicle node->Is used for the transmission rate of (a),representing user +.>Subtask->Transmission data size, +.>Representing user +.>Subtask->The amount of calculation required; user->Subtask->Rental utility generated by rental vehicle virtual computing resources>Expressed as:
Wherein, the liquid crystal display device comprises a liquid crystal display device,representing a vehicle node computing resource lease price;
step 3.3.4, remote cloud node unloading mode:
the cloud can process a plurality of subtasks simultaneously, and the completion time delay of the subtasks to be unloaded to the remote cloud nodeThe method comprises the following steps:
wherein, the liquid crystal display device comprises a liquid crystal display device,representing air edge nodes and users->Cover factor of->Representing the size of the transmitted data +.>Representing ∈k from the air edge node>Uplink transmission rate to base station, +.>Representing user +.>Subtask->Uplink transmission rate to base station, +.>Indicating uplink transmission rate between base station and cloud, < >>Representing cloud allocation to users->SubtasksIs proportional to the computing resource of->Representing remaining computing resources of the remote cloud node; user->Subtask->Leasing utility generated by leasing remote cloud computing resources>Expressed as:
wherein, the liquid crystal display device comprises a liquid crystal display device,representing time slot->Calculating the lease price of the resource by the remote cloud node;
step 3.4, constructing a content acquisition model; the method comprises the following steps:
content acquisition requirements where there may be different access preferences for some users while considering user task offloading, content requirement mobile users can choose to acquire preferred content from vehicle nodes, air edge nodes, and remote cloud nodes, to have、/>、/>Three different acquisition modes of a vehicle node, an air edge node and a remote cloud node are respectively represented;
Step 3.4.1, vehicle node acquisition mode:
wherein, the liquid crystal display device comprises a liquid crystal display device,representing the user of the demand +.>Size of requested content->Representing the user of the demand +.>And mobile vehicle node->Downlink transmission rate between the two; rental utility of requiring a user to download vehicle node content>Expressed as:
wherein, the liquid crystal display device comprises a liquid crystal display device,rental price indicating that the demand user obtains the desired content from the vehicle node,/->Representing the user of the demand +.>Acquiring the preference degree of the content;
step 3.4.2, air edge node acquisition mode:
wherein, the liquid crystal display device comprises a liquid crystal display device,representing the user's need for content retrieval>And air edge node->Is associated with (a) factor(s)>Representing the air edge node +.>Is +.>Downlink transmission rate between the two; rental utility of requiring users to download over-the-air edge node content>Expressed as:
wherein, the liquid crystal display device comprises a liquid crystal display device,representing lease unit price for requiring a user to acquire content from an aerial edge node;
step 3.4.3, remote cloud node acquisition mode:
wherein, the liquid crystal display device comprises a liquid crystal display device,and->Representing cloud node to base station and base station to demand user +. >Is a downlink transmission rate of (a); for demand useHouse->Rental utility of downloading remote cloud node content +.>Expressed as:
wherein, the liquid crystal display device comprises a liquid crystal display device,representing time slot->Requiring the user to acquire the lease price of the content from the remote node;
step 3.5, establishing a system utility function weighted by multiple performance indexes; the specific process is as follows:
step 3.5.1, time slots are allocated by taking into account the different possible offloading modesDelay and +.>Expressed as:
wherein, the liquid crystal display device comprises a liquid crystal display device,representing subtasks->Dependency of attribute factor->Is->Dependency variable value at attribute time, +.>Representing the total amount of subtask dependent properties +.>Representing user +.>Subtask->Selecting self-unloading->Representing subtasks->Delay in the completion of the self-offloading of the user node +.>Representing user +.>Subtask->Selecting over-the-air edge node offload,>representing subtasks->Offloading to an over-the-air edge node->Completion delay of->Representing user +.>Subtask->Selecting vehicle virtual resource sharing offload, +.>Task completion latency representing offloading of subtasks to vehicle virtual resource sharing +.>Representing user +.>Subtask->Selecting remote cloud for task offloading, +.>Representing the completion time delay of the subtask offloading to the remote cloud node;
wherein, the liquid crystal display device comprises a liquid crystal display device,representing user +.>Subtask->Leasing the air edge node computing resource generated leasing utility,representing user +.>Subtask->Rental utility generated by rental vehicle virtual computing resources, +.>Representing user +.>Subtask->Leasing the leasing utility generated by the remote cloud computing resource;
step 3.5.3, combining different content acquisition strategies, requiring the user to be in the time slotContent acquisition delay and->Expressed as:
wherein, the liquid crystal display device comprises a liquid crystal display device,for indicating demandHouse->Selecting a vehicle node content acquisition mode,/->Representing the user of the demand +.>Selecting an over-the-air edge node content acquisition mode, +.>Representing the user of the demand +.>Selecting a remote cloud node content acquisition mode;
step 3.5.4, obtaining service lease utility of edge node content by user associated with user demand content size and popularityExpressed as:
wherein, the liquid crystal display device comprises a liquid crystal display device,rental utility indicating that the user is required to download the contents of the vehicle node,/->Rental utility indicating that the user is required to download the content of the edge node in air,/->Representing leasing utility of a demand user to download remote cloud node content;
step 3.5.5, quantitatively analyzing, calculating and unloading and obtaining time delay of contentSystem utility function for weighted sum representation of payout rental utility of user rental edge resources and payout rental utility of content acquisition service :
Wherein, the liquid crystal display device comprises a liquid crystal display device,、/>、/>、/>respectively represent time slots->Delay and +.>Computing resource lease utility generated by lease edge node>The demand user is in time slot->Content acquisition delay and->Service rental utility of user obtaining edge node content>Weight coefficient of>、/>Respectively representing computing resource lease utility and obtaining discount factors of edge node content service lease utility;
and 4, constructing an air-ground collaborative unloading and content acquisition optimal strategy based on edge service knowledge graph perception for different types of user demands by adopting a theoretical optimization and clipping near-end strategy optimization algorithm.
2. The method for air-ground collaboration offloading and content acquisition based on knowledge-graph perception according to claim 1, wherein in the step 1, an edge city air-ground collaboration scene in which a 5G macro base station covers remote clouds, air edge nodes, road moving vehicles and two types of mobile users with different types and different requirements coexist is considered; the macro base station is provided with an edge knowledge server for knowledge spectrum sensing, senses physical entity information of users, vehicles and aerial edge nodes in a coverage area, is connected with a remote cloud in an optical fiber communication mode, and contains a large amount of computing resources and cache content fragment sets; meanwhile, quantitatively labeling the positions of different physical entities in the scene by establishing a three-dimensional coordinate system; the unmanned aerial vehicle with fixed height flies at a non-preset angle to be used as an air edge node for flexible perception decision deployment, a minimum safety distance is set between adjacent air edge nodes, and the air edge node contains computing resources and part of the existing cache content set updated periodically so as to assist a user to carry out task unloading, relay forwarding and cache content providing, wherein the air edge node and a macro base station communicate through an air link; two kinds of task users with different moving speeds and tolerant time delays exist in the scene, namely a common user and a vehicle-mounted user; a plurality of moving vehicle nodes exist on the surrounding roads of the cell, each vehicle node carries a virtual machine capable of carrying out logic resource migration and an existing cache content fragment set of the vehicle node which is updated periodically, and the virtual machine has vehicle residual computing resources with heterogeneous different sizes and trust dependence social attributes between users and the vehicle nodes; a small number of non-trusted vehicles exist, the vehicles are divided into a host vehicle and auxiliary shared vehicles sharing resources in the edge service process, communication availability ranges exist among different vehicle nodes and among vehicle nodes and task users, and the users communicate with the aerial edge nodes, the vehicle nodes and the host vehicle and the auxiliary shared vehicles through wireless communication links; the air edge node and the vehicle node also serve as edge cache nodes to provide content for users with content acquisition requirements.
3. The space-to-ground collaboration unloading and content acquisition method based on knowledge graph perception according to claim 1, wherein in the step 2, the specific process of constructing the edge service knowledge graph is as follows:
step 2.1, abstracting the air-ground collaboration unloading and content acquisition model network entity constructed in the step 1 into nodes and establishing different logic layers, namely a program task layer, a user layer, a vehicle node layer, an air edge node layer and a cloud node layer;
step 2.2, extracting and analyzing characteristic knowledge of nodes in each layer, and extracting multi-time-varying parameter knowledge relations among the nodes in the layers;
step 2.3, node attribute embedding is carried out on the intra-layer nodes, time-varying parameter knowledge relation arrangement among the nodes is carried out to construct edge relations and parameter weight factors, and multi-time-varying parameter relation arrangement is carried out on the inter-layer nodes according to different edge relation criteria to construct inter-layer node knowledge relations and parameter weight factors;
step 2.4, establishing an undirected weighted graph to obtain an edge service knowledge graph; the method comprises the following steps: the edge service knowledge graph is obtained through dynamic arrangement of different side relation criteria on multi-time-varying parameter knowledge relations and is expressed as an undirected weighted graphThe method comprises the steps of carrying out a first treatment on the surface of the Wherein (1) >Representing node set,/->Representing edges in the undirected weighted graph, +.>Representing the embedded parametric weight.
4. The space-to-ground collaboration unloading and content acquisition method based on knowledge-graph perception according to claim 1, wherein in the step 4, the specific process of acquiring the optimal strategy is as follows:
step 4.1, constructing an optimization problem; the method comprises the following steps:
minimizing computing unloading delay, content acquisition delay of edge service, leasing edge computing resources by users and leasing edge content utility expenditure and achieving the goal of maximizing long-term network utility; the original optimization problem established is expressed as:
Wherein, the liquid crystal display device comprises a liquid crystal display device,represents the maximum number of time slots,/->Representing a non-preset angle; />Representing user +.>Subtasks of->Selecting a user self, an air edge node, vehicle virtual resource sharing and remote cloud node computing unloading mode and requiring users +.>Selecting three content acquisition modes of a vehicle node, an aerial edge node and a remote cloud node; />Indicating that a user can only select one mode to perform task unloading and content acquisition; />Subtask representing user->When different computing task unloading is selected, only one air edge node and one vehicle node are selected as a host vehicle, and a user is required to be +. >When content acquisition is carried out through the aerial edge node, only one node is selected; />Indicating that the distance between the air edge nodes is not less than the minimum safe distance, and the non-preset angle change range is +.>;/>Indicating that the mobile positions of the air edge node, the vehicle node and the user do not exceed the set area limit;a scale factor representing the allocation of computing resources; />And->Indicating that the distribution of computing resources does not exceed the total computing resources of the user node, the air edge node and the remote cloud node during unloading respectively; />Indicating that the task calculation unloading time delay does not exceed the maximum unloading tolerance time delay, and the user content request time delay is smaller than or equal to the maximum unloading tolerance time delay of the content request; />Representing the user's need for content retrieval>And air edge node->Is a correlation factor of (2); />Representing the air edge node +.>And another air edge node->A distance therebetween; />Representing a minimum security distance between adjacent air edge nodes; />Representing the air edge node +.>Is a horizontal axis coordinate value; />Representing user +.>Is a horizontal axis coordinate value; />Representing a mobile vehicle node +.>Is a horizontal axis coordinate value; />A horizontal axis coordinate value representing the coverage area boundary of the macro base station; />Representing the air edge node +.>Is a vertical axis coordinate value of (2); Representing user +.>Is a vertical axis coordinate value of (2); />Representing a mobile vehicle node +.>Is a vertical axis coordinate value of (2); />Representing the vertical axis coordinate value of the coverage area boundary of the macro base station; />Representing user +.>Subtasks of->Is used for unloading calculation time delay; />Representing user +.>Subtasks of->Unloading the maximum tolerant delay; />Representing the user of the demand +.>Selecting content acquisition time delay generated by adding different content acquisition modes; />Representing the user of the demand +.>Maximum tolerant time delay of content acquisition;
step 4.2, optimizing the solution: first, the original optimization problemThe medium discrete variable is relaxed to be changed into a continuous interval variable; second, introducing an upper-bound relaxation variable to the maximum nonlinear term in the objective function>It is subjected toConverting into linear term and adding new constraint condition +.>Optimization problem after relaxation->Is->Performing an equivalent solution; the optimization problem is expressed as follows after simplification>:
Wherein, the liquid crystal display device comprises a liquid crystal display device,representing user +.>Subtasks of->Selecting leasing utilities of computing resources generated by adding different unloading modes;representing the user of the demand +.>Selecting content acquisition time delay generated by adding different content acquisition modes; />Representing the user of the demand +.>Selecting a content service lease utility generated by adding different content acquisitions; />Represents a relaxation of four unloading and three content acquisition discrete variables into a continuous variable between 0 and 1,/v >Representing the user +.>The dependency attribute factor in all subtasks is +.>Upper bound relaxation variable constraint of ∈10->Representing subtasks->Dependency of attribute factor->Is->Dependency variable value at attribute time, +.>Representing user +.>The dependency attribute factor in all subtasks is +.>Upper limit relaxation variable of (2);
the optimization problem is decomposed into three sub-components by a block coordinate descent method and a successive approximation algorithmProblems: user computing offload and content acquisition variable sub-questionsSub-problem of calculating the resource allocation proportion variable ≡>And the sub-problem of deployment angle variable of the air node->The method comprises the steps of carrying out a first treatment on the surface of the The decomposition of a specific sub-problem is represented as follows:
step 4.2.1, giving a non-preset angle and calculating a resource allocation proportion strategy, and solving an unloading and content acquisition strategy;
wherein, the liquid crystal display device comprises a liquid crystal display device,a discount factor representing the utility of a lease of a computing resource, +.>Representing a discount factor for obtaining the lease utility of the edge node content service;
step 4.2.2, given task unloading, content acquisition and non-preset angle strategies, solving a calculation resource allocation proportion strategy;
because no variable coupling relation exists between the distribution of the computing resources and the acquisition of the content, the optimization problem is further simplified to obtain the sub-problem Still belonging to the same solution problem;
step 4.2.3, giving task unloading, content acquisition and calculating a resource allocation proportion strategy, and solving an optimal track strategy of the aerial edge node;
optimization problem with respect to post relaxationAnd variable influence analysis, further simplifying the problem into sub-problems with the same solution when solving the trajectory optimization strategy>;
Wherein, the liquid crystal display device comprises a liquid crystal display device,representing a non-preset angle +.>In the range of 0 to->Between (I)>Representing the air edge node +.>And another air edge node->Distance between->Not lower than the minimum safe distance;
finally, a part of optimized variable parameters are given, variable relaxation is carried out by combining Taylor expansion of local points, the non-convex optimization problem is converted into a convex optimization problem, then different sub-problems are solved, and a theoretical optimal boundary solution of the optimization problem is obtained through multiple iteration and set threshold comparison;
step 4.3, optimizing and perceiving decision analysis under deployment and shared resource pooling based on continuous clipping near-end strategies:
Wherein, the liquid crystal display device comprises a liquid crystal display device,representing user +.>Is (are) moved positions>Representing user +.>Is>Representing user +.>Transition probability of->Representing user +.>Is- >Representing the user of the demand +.>Is (are) acquired request>Representing user +.>Task model with task topology knowledge relationship, +.>Representing a mobile vehicle node +.>Is (are) located>Representing a mobile vehicle node +.>Is>Representing a mobile vehicle node +.>Trust dependent social attributes, +.>Representing a mobile vehicle node +.>Left computing resources of->Representing a mobile vehicle node +.>Existing cached content piece set,/->Indicate->Post-deployment location of individual air edge nodes, +.>Representing the speed of movement of the edge node in air, +.>Representing the air edge node +.>Left computing resources of->Representing the air edge node +.>Part of the existing cached content set,/->Indicating macro base station coordinate position,/->Representing remaining computing resources of remote cloud node, +.>Representing a cached content segment set in the cloud;
analyzing a plurality of knowledge factors influencing the service quality performance, arranging the dynamic knowledge relationship to construct an edge service knowledge graph, and expressing the obtained preprocessing knowledge state information as;
Second, define the complex motion space under each time slot asWherein, the method comprises the steps of, wherein,representing user +.>Subtasks of->Different offloading policies of->Representing different content acquisition strategies of a demand user; the set bonus function under each time slot is denoted +. >。
5. The method for air-ground collaborative offloading and content acquisition based on knowledge-graph perception according to claim 4, wherein in step 4.3, the overall flow of the air-ground collaborative offloading and content acquisition based on perceived deployment and shared resource pooling optimized by continuous clipping near-end policy is as follows:
step 4.3.1, constructing an air-ground collaboration unloading and content acquisition model composed of remote cloud, non-preset track air edge nodes, road mobile vehicles and two types of mobile users with different types and different requirements, and initializing parameters;
step 4.3.2, executing a training round, and initializing a training model to obtain an initial state;
step 4.3.3, executing time slot rounds, analyzing a plurality of time-varying parameter relation arrangement of nodes in layers and among layers to construct an edge service knowledge graph, and obtaining knowledge state information;
step 4.3.4, the knowledge server selects an action strategy through a strategy network;
step 4.3.5, the action strategy is put into the environment for execution, and rewards under the current network state, the next network state, the system utility function, the unloading and content acquisition strategy and the stored experience tuples are obtained;
step 4.3.6, judging whether the parameters of the current strategy network need to be updated, if so, entering another strategy network and evaluating the network to carry out training update, otherwise, continuously executing and updating the network state;
4.3.7, if all time slot training is finished, calculating average network rewards, finishing one training round, and initializing a training model;
and 4.3.8, if the training round is finished, obtaining an average network reward and an optimal service strategy, and outputting the space cooperation unloading and content acquisition scheme as an optimal scheme.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310408910.2A CN116208970B (en) | 2023-04-18 | 2023-04-18 | Air-ground collaboration unloading and content acquisition method based on knowledge-graph perception |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310408910.2A CN116208970B (en) | 2023-04-18 | 2023-04-18 | Air-ground collaboration unloading and content acquisition method based on knowledge-graph perception |
Publications (2)
Publication Number | Publication Date |
---|---|
CN116208970A CN116208970A (en) | 2023-06-02 |
CN116208970B true CN116208970B (en) | 2023-07-14 |
Family
ID=86514890
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202310408910.2A Active CN116208970B (en) | 2023-04-18 | 2023-04-18 | Air-ground collaboration unloading and content acquisition method based on knowledge-graph perception |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN116208970B (en) |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111913723A (en) * | 2020-06-15 | 2020-11-10 | 合肥工业大学 | Cloud-edge-end cooperative unloading method and system based on assembly line |
CN112800247A (en) * | 2021-04-09 | 2021-05-14 | 华中科技大学 | Semantic encoding/decoding method, equipment and communication system based on knowledge graph sharing |
WO2021233053A1 (en) * | 2020-05-22 | 2021-11-25 | 华为技术有限公司 | Computing offloading method and communication apparatus |
WO2022121559A1 (en) * | 2020-12-07 | 2022-06-16 | 中兴通讯股份有限公司 | Mobile communication system control method, network management and control agent, system, and storage medium |
Family Cites Families (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112732718B (en) * | 2021-01-06 | 2022-06-14 | 华南理工大学 | Intelligent structured data method, system and medium for cloud side |
CN114915629B (en) * | 2021-02-10 | 2023-08-15 | 中国移动通信有限公司研究院 | Information processing method, device, system, electronic equipment and storage medium |
CN113742046A (en) * | 2021-09-17 | 2021-12-03 | 苏州大学 | Flow grooming cloud-side computing network computing resource balanced scheduling method and system |
CN115129888A (en) * | 2022-06-14 | 2022-09-30 | 大连理工大学 | Active content caching method based on network edge knowledge graph |
CN115243217B (en) * | 2022-07-07 | 2023-07-18 | 中山大学 | DDQN-based terminal Bian Yun cooperative scheduling method and system in Internet of vehicles edge environment |
-
2023
- 2023-04-18 CN CN202310408910.2A patent/CN116208970B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2021233053A1 (en) * | 2020-05-22 | 2021-11-25 | 华为技术有限公司 | Computing offloading method and communication apparatus |
CN111913723A (en) * | 2020-06-15 | 2020-11-10 | 合肥工业大学 | Cloud-edge-end cooperative unloading method and system based on assembly line |
WO2022121559A1 (en) * | 2020-12-07 | 2022-06-16 | 中兴通讯股份有限公司 | Mobile communication system control method, network management and control agent, system, and storage medium |
CN112800247A (en) * | 2021-04-09 | 2021-05-14 | 华中科技大学 | Semantic encoding/decoding method, equipment and communication system based on knowledge graph sharing |
Also Published As
Publication number | Publication date |
---|---|
CN116208970A (en) | 2023-06-02 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Hu et al. | Twin-timescale artificial intelligence aided mobility-aware edge caching and computing in vehicular networks | |
CN113282368B (en) | Edge computing resource scheduling method for substation inspection | |
Chen et al. | Multiuser computation offloading and resource allocation for cloud–edge heterogeneous network | |
Wu et al. | Deep reinforcement learning-based computation offloading for 5G vehicle-aware multi-access edge computing network | |
CN111182495B (en) | 5G internet of vehicles partial calculation unloading method | |
CN111132191A (en) | Method for unloading, caching and resource allocation of joint tasks of mobile edge computing server | |
CN112512065B (en) | Method for unloading and migrating under mobile awareness in small cell network supporting MEC | |
Wang et al. | Q‐learning based computation offloading for multi‐UAV‐enabled cloud‐edge computing networks | |
Lin et al. | Deep reinforcement learning-based task scheduling and resource allocation for NOMA-MEC in Industrial Internet of Things | |
CN114698128A (en) | Anti-interference channel selection method and system for cognitive satellite-ground network | |
Tan et al. | Resource allocation of fog radio access network based on deep reinforcement learning | |
Li et al. | Multi-task offloading scheme for UAV-enabled fog computing networks | |
Hu et al. | Deep reinforcement learning for task offloading in edge computing assisted power IoT | |
Tefera et al. | Congestion‐aware adaptive decentralised computation offloading and caching for multi‐access edge computing networks | |
KR101924628B1 (en) | Apparatus and Method for controlling traffic offloading | |
Li et al. | Collaborative optimization of edge-cloud computation offloading in internet of vehicles | |
CN116208970B (en) | Air-ground collaboration unloading and content acquisition method based on knowledge-graph perception | |
CN115361453B (en) | Load fair unloading and migration method for edge service network | |
Zhu et al. | Computing offloading decision based on multi-objective immune algorithm in mobile edge computing scenario | |
CN114615730B (en) | Power distribution method for content coverage of backhaul-limited dense wireless network | |
CN116321293A (en) | Edge computing unloading and resource allocation method based on multi-agent reinforcement learning | |
CN115766478A (en) | Unloading method of air-ground cooperative edge computing server | |
Lei et al. | Partially collaborative edge caching based on federated deep reinforcement learning | |
CN115052262A (en) | Potential game-based vehicle networking computing unloading and power optimization method | |
CN114626298A (en) | State updating method for efficient caching and task unloading in unmanned aerial vehicle-assisted Internet of vehicles |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |